With AI, building is no longer the hard part

A rota tool, in two days. A planning and time-tracking tool for 35 people, in a week.

Building is no longer what holds you back. What holds you back is getting someone to use it.

What you can build in days now

Three examples I know well, because I built them myself. All three are in use.

Mia's Care is a nail salon in Madrid. Its tool for shifts, holidays and time off was built with AI in two days. It costs €0 a month.

Lacia OnTrack is the internal tool of a packaging design agency: planning, logged hours and the real cost of each project. I built it on my own, with AI, in a week, with iterations. It has 35 users and is still in use.

The wedding website for some friends is three products on the same codebase. A website to RSVP, QR tools for the wedding day and a panel for the couple.

And they really work: the Mia's Care tool checks overlaps, notice periods and minimum cover in the database. Lacia OnTrack suggests which designer is the best fit for each new project. The wedding panel has a table editor where you drag each guest to their seat.

The ones that work already had users

All three have something else in common: their users existed before the first line of code.

At Mia's Care, four people who asked for their shifts over WhatsApp and face to face.

For Lacia OnTrack, 35 people at an agency that ran its strategic projects in Excel.

At the wedding, about 120 guests. Half have already confirmed through the website.

They were requests: the users came with the brief. Nobody had to go out and find them. And at the salon and the agency, nobody had to be convinced the problem existed either. They already had it.

Distribution was solved before building. The day each tool was ready, it was already clear who would get it and what for.

That's why they get used. Not because of how fast they were built.

Having the users in front of you from day one gives you something else too: you know what to change next. The Mia's Care tool keeps growing with whatever the salon needs.

Where it does get hard: being found

ListoChina is the opposite case. It's under construction: a web app for Spanish speakers who want to go to China or understand its culture. It's for planning the trip with AI, reading about culture and travel tips, and learning Chinese.

I'm building it because it's the app I want to use. The web app is already usable and still a work in progress.

There's no request here, and no team of four waiting for the tool. There's an audience: people thinking about travelling to China. You don't know their names or where they are. And, to begin with, they don't know the app exists.

Here, building is the easy part. The hard part is going out to find them: being found when they're planning the trip. And getting them to come back.

Being your own first user helps you know what to build. But one user isn't a market.

Sell before you build

When nobody brings you the job, you have to win it before you build. In B2B, that's how it already worked before AI.

At WinView, there were signed clients before the product existed. We ran discovery with them, looking for what they had in common. That led to a base product that reached more than 20 clients in production.

At acens, inside a large company, the order was the same, just in a different form. Every product started with its business case and its go-to-market plan, defended before senior management.

The order matters. First, who will use it and how it reaches them. Then, the product.

When building took months, that order took care of itself: putting a team to work on code without knowing who it was for was too expensive. Now you can skip it and have something working in a week.

That's the new risk. Not that the MVP turns out badly. That it turns out well and nobody uses it.

What to do before you open the editor

With AI, the temptation is to open the editor first. It costs little, and in an afternoon you can already see something on screen.

Before you ask the AI for the first screen, answer three questions. In writing.

1. Who are your first users? It can be just four, like at the salon. But by name. If you can't come up with a single one, you don't know who it's for yet.

2. How will they find out about it? A message, a meeting, a link in the place where they already are. "I'll share it on social media" isn't an answer.

3. What has to happen for them to use it in the second week? Anyone will try it once. Coming back is a different matter.

If you can't answer, the MVP can wait. It'll still only take you days.

If you can answer, build it now. Don't chase the perfect version: just enough for those first users to come back.

Code is no longer the excuse. Get to work on distribution.

Let's go.